Validation statistics and Jack-knife analysis of the variable contributions.

*Values are significantly different from random prediction (One-Sample t-test, p < 0.01, One-Sample Signed Rank test, p < 0.03 if normality failed).
Values for the regularized training gain of the Jack-knife test indicates the variable contribution to the model, the higher the value the greater the contribution (values are not directly comparable between the different sites). Standard deviations in parentheses.